When I dabbled in the webpage parsing space, for a project that allowed researchers to undertake content analysis of student web content. Reliably parsing the content, such as identifying each whole comment on a page was the difficult issue. I'm not of Hypothesis ability in this area, as I've just signed up to make this annotation, but if you've solved that problem then I can imagine a whole range of interesting apps might be created as plugins for hypothesis.